Deploy and Host Vane (AI Answering Engine) with SearXNG on Railway

Vane, formerly Perplexica, is an open-source, privacy-focused AI answering engine. It searches the web through SearXNG, reads the results with the LLM of your choice, and returns cited answers in a Perplexity-style chat UI. It supports OpenAI, Anthropic, Google Gemini, Groq, Ollama, LM Studio, and any OpenAI-compatible server, plus local embeddings that need no API key.

About Hosting Vane

Vane is a Next.js app that keeps its settings, chat history, and uploads in a single data directory, and it needs a SearXNG instance with the JSON format enabled. Upstream's default Docker image bundles SearXNG inside the Vane container. This template follows upstream's documented alternative: the official slim image (itzcrazykns1337/vane:slim-v1.12.2) talks to a separate, private SearXNG service over Railway's internal network. Vane gets a public domain, a healthcheck on /, and a volume at /home/vane/data. SearXNG is built from the same source as the published OpenClaw and Hermes templates (JSON API on, rate limiter off), has no public domain, and keeps its settings.yml on its own volume.

Common Use Cases

Private Perplexity alternative: ask questions and get cited answers without sending queries to a hosted search product
Research assistant: Speed, Balanced, and Quality modes plus academic sources (arXiv, Google Scholar, PubMed)
Search API for your apps: POST /api/search returns an answer with sources for any query
Document Q&A: upload PDFs, text, and images and ask questions about them
Local-model front end: point it at Ollama or LM Studio and keep everything self-hosted

Dependencies for Vane Hosting

SearXNG with json in search.formats (included, private only)
One LLM provider: an API key for OpenAI, Anthropic, Gemini, or Groq, or a reachable Ollama / LM Studio / Lemonade / OpenAI-compatible endpoint. Not needed to boot; added in the setup wizard
Embeddings: the built-in Transformers provider runs locally with no key, or use OpenAI / Gemini embeddings

Deployment Dependencies

Implementation Details

First steps after deploy:

Open the Vane domain. The setup wizard runs on first boot.
Add a model provider and API key, then choose a chat model and an embedding model (the Transformers option needs no key).
The SearXNG URL is pre-filled from SEARXNG_API_URL. Finish the wizard and ask your first question.
Change providers, models, or the SearXNG URL later in Settings. Env vars are only read on first boot; the Settings page is the source of truth afterwards.

Key variables:

# Vane
PORT=3000
HOSTNAME=0.0.0.0
SEARXNG_API_URL=http://${{SearXNG.RAILWAY_PRIVATE_DOMAIN}}:${{SearXNG.PORT}}
OPENAI_API_KEY=            # optional seed; also ANTHROPIC_API_KEY, GEMINI_API_KEY, GROQ_API_KEY

# SearXNG (private only)
PORT=8080
SEARXNG_BASE_URL=http://${{RAILWAY_PRIVATE_DOMAIN}}:8080/
SEARXNG_SECRET_KEY=${{ secret(32, "abcdefgABCDEFG123456") }}
ENABLE_ALPINE_PRIVATE_NETWORKING=true

Vane has no built-in login. Anyone with the URL can chat and read settings through /api/config, so keep the URL private or place an authenticating proxy in front of it. The pinned slim image omits Playwright, so Quality mode uses search snippets instead of full page reads; switch the image to itzcrazykns1337/vane:slim-latest if you want page scraping.

Why Deploy Vane on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying Vane on Railway, you get a working AI answering engine with its own private search backend, generated secrets, persistent storage, health checks, managed SSL, and private networking, ready for your API keys in minutes.